Introduction

In an automotive electronic component EMC test, a cable bundle with a pre-defined length is required to connect the Device Under Test (DUT) with the rest of the test system, e.g. battery, load simulator, etc. Such a long cable introduces resonances in the system under test, causing radiation and immunity problems at the resonance frequencies. Some of the automotive EMC test scenarios give thought to this, they require multiple test locations on the wiring harness in order not to miss the antinodes of standing waves, for example, CISPR-25 conducted emission – current probe method claims the measurement shall be done separately with the current clamp positioned 50 mm and 750 mm from the DUT; ISO-11452-4 claims the Bulk Current Injection test shall be performed with the injection probe placed 150 mm, 450 mm, 750 mm from the DUT.  

But compared with “how not to omit the resonance”, an EMC engineer is usually more interested in how to predict the resonance behavior of the system design and how to improve in case of test failures. This article will start with some typical failure cases in automotive EMC tests in which the cable resonances play dominant roles; then the mechanism of the cable resonance is discussed, where further test examples and simulation results are shown to demonstrate the root causes of such resonance; finally, system design details are provided that impact the cable resonance behavior.

Typical Cable Resonance Behaviors

A typical phenomenon of cable resonance is the wideband emission peaks in certain frequency range. If an EMC engineer finds his Radiated Emission (RE) or Conducted Current Emission (CCE) measurements gratuitously peaks in a range that will be discussed later, while they cannot find any similar peaks in the PCB near-field scan, they should consider the possibility of cable resonance.

These resonance peaks are frequently observed in components comprising automotive Ethernet or video Serdes connections, whereas these two communication protocols are quite common in vehicle domain controllers. To begin, the article will examine a Driver Monitor System (DMS) component example, which contains a 100Base-T1 channel for connection with the domain controller and a GMSL2 channel for connection with a camera. Therefore, the EMC test system of the DMS comprises the DUT itself, an Ethernet load, a camera load, and a 1.8m wiring harness. Figure 1a shows the CCE test result of the DUT system. Figure 1b shows the Common Mode (CM) emission of the 100Base-T1 IOs. Figure 1c shows the power spectrum distribution (PSD) of the 100Base-T1 differential data. The CCE test setup contains a 1.8 m long cable, while the CM emission and PSD were directly measured at the connector of the component, without the presence of long cable.^1,2,3 

Figure 1. (a) CCE test result of the DUT, (b) CM emission test result of the DUT’s 100Base-T1 signal, and (c) PSD test result of the DUT’s 100Base-T1 signal.

In Figure 1a, a distinct peak at 33.5 MHz can be observed, which causes a local failure at that frequency. This peak could disappear if the 100Base-T1 connection is unlinked, which correlates the problem to the automotive Ethernet part. However, Figures 1b and 1c show that the automotive Ethernet signal had CM and DM spectra covering that frequency range, but no comparably sharp peak could be found. The PCB near-field scan also did not report any similar peaks in that frequency range. In addition, some trial measurements showed that the grounding state of the DUT and camera could significantly affect the amplitude and frequency of this peak. These phenomena implied the importance of the cable system.

Figure 1 is a relatively simple example. The emission peaks caused by cable resonance may behave diversely depending on the DUT system configuration. This article will cover more conditions following the discussion of the cable resonance mechanism.

Mechanism of the Cable Resonance

In an automotive EMC test environment, the system is placed 50 mm above the reference ground plane. The DUT and load may be grounded or floating, depending on the test plan. For instance, when one side is grounded and the other is floating, the cable system is analogous to a folded monopole antenna, whose primary resonant frequency would be where the 1/4 wavelength matches the cable length. Similarly, higher-order harmonics at higher frequencies, including third, fifth, etc., are also predictable. 

It should be mentioned that the “monopole antenna” here is lying horizontally instead of standing vertically above the reference ground plane, which produces some more parasitic capacitance to the system, and this will in turn make the primary resonant frequency smaller than estimated. Besides, the 50 mm-thickness supporting material’s relative permittivity (ε_r ≤ 1.4 is accepted by CISPR-25) also contributes to lowering the resonant frequency. In addition, when a physically large DUT/load is present on the floating side, the frequency will further decrease due to the large parasitic capacitance imposed on the reference ground plane. However, the λ/4 estimation method is already sufficient for EMC risk evaluation. In a typical emission test setup, the cable length should be 1.7 m to 2 m, as stated by CISPR-25; in practice, a 1.8 m length is usually adopted because it factors in both convenience of operation and reusability in other EMC test items. In this case, the primary resonant frequency can be estimated as f₀ = c/λ = c/(4 × 1.8 m) ≈ 41 MHz. For our test experience in various automotive components, the measured resonant peaks vary from 30 MHz to 40 MHz. Considering those reasons for the extra capacitance, this shows satisfactory agreement with the λ/4 estimation.

Many automotive component EMC test systems, whether intentional or unintentional, contain this type of single-end grounded wire. For example, if a DUT with a metal shell has a camera load, the coaxial cable screen of the video SerDes channel is usually directly connected to the metal shell of the DUT and the camera. If the EMC test system plans to locally ground the DUT while floating the camera on the other side, one will find that the coaxial cable screen is single-end grounded, together with a 30 to 40 MHz emission peak.  This is the same for automotive Ethernet channels using a Shielded Twisted Pair (STP), for example, 1000Base-T1, where the cable shield can also serve as a resonator. Other possibilities may be ground wires or signal IO wires. Depending on the component and test system design, whenever an arbitrary wire has low impedance to the reference ground plane on one side and high impedance on the other side, awareness of such a resonant structure should be raised.

The presence of a resonant structure does not necessarily induce EMI failures, while other essentials are the electromagnetic power source in that resonant frequency range and the coupling path between the source and the resonant structure. For example, in the previous DMS example, 100Base-T1 channel had abundant spectral content at 30-40 MHz. It was using Unshielded Twisted Pair, and was sharing a multi-pin connector with the video coaxial cable (obsolete design), which provided ample coupling from the 100Base-T1 channel to the resonant coaxial cable screen. Apart from the communication signal, switched power supplies can also couple the power noise into the resonant structure if not properly shielded, which is shown in another example later in this article. 

Communication Channel Signal as Noise Source

In the previous DMS example, the coax served as a resonant structure, and was excited by external 100Base-T1 channel signal through near-field coupling. It is worth mentioning that the signal or noise propagating inside the coax or STP can also excite the resonating screen. This is because of the limited shielding effectiveness (SE) of the cable screen. The SE, or alternatively, the transfer impedance of a cable screen, depends on the cable manufacturer’s braiding and foil design. For simplicity, the transfer impedance Zt can be modeled using the transfer resistance (Ω/m) and transfer inductance (H/m). For a typical RG58 cable screen, the two parameters are 0.0147 Ω/m and -6.6e-10 H/m respectively.

The effect of transfer impedance can be verified in a simulation example in CST Studio Suite.⁴ In Figure 2, a 1.8 m coax connects the locally grounded DUT to the floating camera. A 350mVAC signal was stimulated on the DUT side. The characteristic impedance and termination of the coax channel on both sides are 50 Ω. The CCE was calculated by directly integrating the (magnetic field) along a closed trace encircling the cable.

In Figure 3, by halving and doubling Z_t, 6 dB variations in the CCE results can be observed. This implies that Z_t (or alternatively, SE,) dominates the coupling mechanism in this example. In engineering practice, it would be helpful to choose a cable with better SE to reduce this type of coupling.

Figure 2. Schematic of the simulated test setup. 

Figure 3. CCE simulation results for different values of Z_t, where Z_t0 denotes the transfer impedance of a typical RG58 cable screen.

Besides the cable screen SE, an EMC engineer should pay special attention to the connector design of the cable harness. Some low-cost connector tips were designed with incomplete shielding, and could thus become a dominant coupling path. This is independent of the cable screen’s nominal shielding property, so a low-quality connector can easily destroy the cable assembly’s overall performance even with a faultless cable screen. Figure 4 shows two Fakra connector tips, one with inadequate shielding and the other without. The former is prone to failure during the EMC test.

Figure 4. (a) Fakra connector tip with inadequate shielding vs. (b) well-designed Fakra connector tips.

For differential channels, for example, in the case of STP, the imbalance of the differential channel is another key factor that contributes to excess EMI. Sometimes, an EMC engineer may find that his emission test data change remarkably by simply replacing the cable assembly with another one. This can be attributed to the cable assembly manufacturer’s quality control of differential channel symmetry. The asymmetry partly converts differential mode (DM) signal into CM, and for CM signals, it is analogous to the former coax example. The asymmetry can be quantified using SDC parameters measured by a 4-channel VNA. Some recommended SDC limits for automotive STP cable assemblies are available.^5,6 The channel performance shall be verified before the EMC test. 

As a negative example from our project experience, Figure 5 shows an 1000Base-T1 connector of a vehicle domain controller. The plastic fixture can hold two metal clips in each STP channel, so the screen can be grounded to PCB symmetrically, while only one clip was installed by the manufacturer. Omitting a grounding clip in this way seldom affects the data link quality, as SDDS21 will hardly change within the signal bandwidth. However, this strong asymmetry in the connector region will boost SDC and, therefore, EMI. This domain controller passed EMI regulation after fixing this symmetry problem.

Figure 5. An EMC-unfriendly connector example.

Nearfield Coupling as Noise Source

Besides the communication channel spectrum mentioned above, power noise on the PCB is another potential noise source that can couple into the resonant structure. In our project experience, a camera product was originally designed with aluminum shell and had passed all EMI regulations. Later, the shell was re-designed with plastic material for cost reduction reasons, and the cable resonance peak became prominent again. Some measurement data shown below reproduce this phenomenon and demonstrate the noise source and coupling path.

In Figure 6a, the CCE measurement was performed without any modification of the plastic shell. In Figure 6b, a copper foil with electrical connection with the PCB ground was applied to wrap the plastic shell. However, in this case, the copper foil did not touch the cable or connector’s screen directly. In Figure 6c, the copper foil fully encloses the component by wrapping the coax screen in the vicinity of the connector. The measurement results of these three setups are shown in Figure 7. In the first case, the primary resonant frequency was approximately 37 MHz, which agreed well with the previous analysis. In the second case, inadequate shielding resulted in minor reduction in the emission peak. The resonant frequency also decreased due to the copper foil’s extra capacitance to ground. In the last case, the thorough shielding provided a conspicuous reduction in the emission peak, and EMI regulations were also satisfied in this way. Third harmonics are also observable in Figure 7, which also shows a similar trend of improvement in the presence of the copper foil enclosure.

Figure 6. (a) CCE test setup on the original camera, (b) inadequate copper foil wrapping, and (c) copper foil fully enclosed the component and connector.

Figure 7. The measurement results for the three conditions shown in Figure 6.

The principle of the difference between the three conditions can be explained by the difference in the noise return paths. In Figure 8a, the power noise source on the PCB around the connector can draw a common-mode current onto the outer surface of the coax through capacitive coupling. The coupled current returned to the PCB mainly through the red dashed line. This corresponds to the condition in Figure 6a. In Figure 8b, because the shielding in Figure 6b did not fully enclose the DUT (at the coax connector region), the current return path did not essentially change. Therefore, the result in Figure 7 showed little improvement. In Figure 8c, owing to the complete shielding in Figure 6c, the noise return path became much smaller, and no coupled current entered the monitoring region of the current clamp.

Figure 8. Schematic representations of the noise return path under the three conditions shown in Figure 6.

To reduce this coupling without using a metal case or shield in a component, an EMC designer should figure out the potential coupling noise source and place them far from the connector region, or to the opposite side of the PCB.  

Grounding Condition Dominates the Resonance Behavior

Depending on the grounding design of the two sides of the cable, the resonant frequencies can differ significantly. This is because the equivalent model of the resonant structure changes according to the cable grounding configuration. Figure 9 compares the results of three typical cases, including the simulated CCE results and the field distribution. In this example, a 500 mV differential signal was propagating in the communication channel, and the STP differential pair was designed with a 0.5 nH mismatch. This asymmetry caused DM to CM conversion, and the CM return current flowing on the STP shield exposed the cable shield resonance problem. In the leftmost column, the cable system was grounded on one single end, and the CCE result was analogous to that in Figure 3. The primary resonant frequency here was 29 MHz, slightly lower than 33 MHz in Figure 3, owing to the larger component size and, therefore, larger parasitic capacitance in this example. The side view of E field distribution is shown at the bottom, with its node at the grounded end and antinode at the floating end.

In the latter two cases, the cable system was floating or grounded on both ends. When resonance occurs, the system can be split into two identical single-end grounded structures mirroring each other, thus, the primary resonant frequency can be estimated by doubling the result of the λ/4 estimation. In the middle column, the cable system was floating on both ends. The E field distribution resembled that of a dipole antenna. The primary resonant frequency was 46 MHz, lower than twice the result of the λ/4 estimation (2×41 MHz). This is because the physically large components (15 cm × 15 cm in this simulation) were located in the high E region on both ends, and thus effectively contributed extra parasitic capacitance to the resonant structure and therefore reduced the resonant frequency. In the rightmost column, the cable system was grounded on both ends, and the E field distribution showed nulls on both ends. In this way, the resonant system will not “feel” much effective capacitance from the physically large components. The primary resonant frequency here was 82 MHz, which corresponded quite well with twice the frequency of the λ/4 estimation.

Figure 9. Three grounding configurations and their CCE and E field distributions (side view).

The three basic models in Figure 9 clarify that the grounding condition dominates the resonance behavior. Higher-order harmonics also vary in response to the grounding condition and can be analyzed in a similar manner. However, in engineering practice, these three basic models are oversimplified. There are far more than barely one Coax/STP in a component’s cable bundle, so more than one basic model in Figure 9 may appear in a component’s test system. Each of these wires exhibits unique termination impedance, and all of them are tightly coupled. For these reasons, the overall resonant frequency and amplitude of the system are usually difficult to predict precisely without meticulous modeling and simulation. In the opinion of the authors, compared to adding complexity to the simulation, it is more time-efficient to keep the mechanism of the resonance in mind and know the ways to mitigate it.

Conclusion

This article studies a cable resonance problem in automotive component EMC tests, which usually induces wideband emission peaks in predictable frequency ranges. For a single-end grounded 1.8 m cable system, the problematic frequency is typically 30 to 40 MHz, corresponding to a λ/4 estimation. The problematic frequency may exhibit more diversity by involving different grounding conditions, multiple wire couplings, high-order harmonics.

In general, there are two ways to improve. One method is to optimize the resonant structure. For example, adjusting the termination/grounding conditions of the two ends can achieve an immediate modification in the resonant frequency.

The second approach is to reduce the power coupling into the resonant structure. The noise sources are usually communication channels or power supplies that have spectrum in the resonant frequency range. Some key points that worth checking are listed below.

1. SE of the shielded signal channel. Apart from the nominal shielding property of the cable screen, special attention should be paid to the connector design. A low-quality connector can significantly degrade the overall SE.
2. Symmetry of the entire differential signal channel. This can be quantified using SDC measurements with VNA.
3. Power supply noise near the connector. Try to improve the PCB layout or add a shield to that power supply.

REFERENCES

1. CISPR 25, “Vehicles, boats and internal combustion engines - Radio disturbance characteristics - Limits and methods of measurement for the protection of on-board receivers,” IEC
2. B. Körber, “IEEE 100BASE-T1 EMC Test Specification for Transceivers,” FTZ Zwickau, Open Alliance, June 20, 2020.
3.  C. Donahue, “IEEE 100BASE-T1 Physical Media Attachment Test Suite,” UNH-IO, Open Alliance, June 6, 2017.
4. Dassault Systémes, “CST studio suite - electromagnetic field simulation software.” Accessed: Oct. 5, 2023.
5. "IEEE Standard for Ethernet Amendment 4: Physical Layer Specifications and Management Parameters for 1 Gb/s Operation over a Single Twisted-Pair Copper Cable," in IEEE Std 802.3bp-2016 (Amendment to IEEE Std 802.3-2015 as amended by IEEE Std 802.3bw-2015, IEEE Std 802.3by-2016, and IEEE Std 802.3bq-2016), vol., no., pp.1-211, 20 Sept. 2016, doi: 10.1109/IEEESTD.2016.7564011.
6. “Channel and Component Requirements for 1000BASE-T1 Link Segment Type A (STP),” Open Alliance, June 24, 2020.